Direct interaction with no correlation: An experimental pitfall in neural systems

Neural systems continuously optimize how organisms process their environment and are highly dynamic. Building predictive models of these systems is challenging due to the large number of their elements. Therefore, in experimental and descriptive neurobiology the researcher typically does not seek to catalogue all variables that affect one another, but rather tries to isolate variables that interact directly. Because of methodological limitations, observed variables are often measured near equilibrium. The presented analysis demonstrates that statistical tests performed on such equilibrium values are fundamentally incapable of detecting direct interactions in a large subset of simple dynamical systems. Some of these problems can be avoided by using explicit statistical models that include time as a variable.

► Neuroscience seeks to discover direct interactions among variables.
► Direct interactions may not be detectable if variables are measured at equilibrium.
► Explicit statistical models with time as a variable can circumvent this problem.